Convert tables and figures in the report to RDF with multimodal LLMs

[wbcbugfree]

The report contains a large number of tables and figures that contain much information not mentioned in the text. In this stage, we focus on converting the text to RDF, but these tables and figures also contain information that is crucial for the soil health knowledge graph. Multimodal LLMs are a feasible solution for extracting this information and converting it to RDF, and will be tested soon.

[robknapen]

I want to try out LlamaParse soon-ish :-) At the moment it scans 1000 pages per day for free. So maybe you can already use on the report and check if results improve.

https://cloud.llamaindex.ai/parse

[wbcbugfree]

Refer to the prompt on processing figures, especially flowcharts and diagrams:

• End To End Multimodal Knowledge Graph Creation from Texts and Images & Querying in Natural Language using LangChain and Neo4j.